Optoelectronic properties and applications of organic semiconductors are defined by the formation and dynamics of excitons which are electron-hole pairs bound electrostatically, with the energy difference to separated charges known as exciton binding energy. The fundamental mechanisms underlying the exciton binding are complex and its ultrafast sub-200fs timescales [1] impose stringent time-resolution requirements on the experimental technique. Previously, ultrafast optical pump-probe (PP) spectroscopy [2,3], two-photon photoemission [4,5], photoluminescence and anisotropy [6,7] has proved to be a useful tool to capture exciton dynamics, showing a rapid creation of excitons with an ultrafast (<200fs) exciton localisation and charge separation. Although these methods track well the population dynamics of states with specific energies, they do not provide direct access to the binding energy of excitonic states and may not distinguish excitons from other excited species. This lack of bound states sensitivity can be overcome using pump-push-photocurrent (PPPC) techniques [8,9]. In PPPC, after the initial excitation of the system by the pump-pulse, a second pulse, that is the push-pulse, re-excites the bound excited states promoting their dissociation and the push-induced photocurrent originated from the re-excited bound excited states is measured.

Here, we apply a combination of pump-probe (PP) and pump-push-photocurrent (PPPC) spectroscopies with sub-10-fs time resolution to separate and track in time the ultrafast formation of bound excitons in a ‘classical’ conjugated polymer, polyfluorene poly(9,9-dioctylfluorene) (PFO), device. We found that excitons created by near-absorption-edge excitation are intrinsically bound states, or are becoming such within 10 fs after excitation. At the same time, hot excitons with >0.3 eV excess energy do have a chance for spontaneous dissociation and acquire fully bound character only at a 50-fs timescale. Excitation fluence dependent measurements show that the exciton binding does not depend on the density of excited states and that exciton-exciton annihilation contributes to charge dissociation by giving excess energy to the bound exciton. Furthermore, we develop a simple global kinetic model which reproduces PP and PPPC experiments with a compact set of shared parameters, supporting the above conclusions. Finally, we use the non-adiabatic excited state molecular dynamics package (NEXMD) [10] to simulate singlet exciton evolution in an oligomeric model of this polymer. Our simulation confirms that binding energy of the near-band-edge S1 excitons does not evolve significantly beyond the ~20-fs timescale and that these near-band-edge states can be populated through hot exciton states cooling with characteristic time of 50 fs.